Flood monitoring system

Sensors 2012, 12, 4213-4236; doi: 10. 3390/s120404213 OPEN ACCESS sensors ISSN 1424-8220 www. mdpi. com/journal/sensors Article A Real-Time Measurement System for Long-Life Flood Monitoring and Warning Applications Rafael Marin-Perez 1, , Javier Garc? a-Pintado 2, 3 and Antonio Skarmeta G? mez 1 ? o 1 Department of Information andCommunicationEngineering, University of Murcia, Campus de Espinardo, E-30100, Murcia, Spain; E-Mail: ska[email protected]es 2 Euromediterranean Water Institute, Campus de Espinardo, E-30100, Murcia, Spain; E-Mail:[email protected]om 3 National Centre for EarthObservation, University of Reading, Harry Pitt Building, 3 Earley Gate, Whiteknights, Reading RG6 6AL, UK Author to whom correspondence should be addressed; E-Mail:[email protected]es. Received: 7 February 2012; in revised form: 14 March 2012 / Accepted: 22 March 2012 / Published: 28 March 2012 Abstract: A ? ood warning system incorporates telemetered rainfall and ? ow/water level data measured at various locations in the catchment area. Real-time accurate data collection is required for this use, and sensor networks improve the system capabilities.

However, existing sensor nodes struggle to satisfy the hydrological requirements in terms of autonomy, sensor hardware compatibility, reliability and long-range communication. We describe the design and development of a real-time measurement system for ? ood monitoring, and its deployment in a ? ash-? ood prone 650 km2 semiarid watershed in Southern Spain. A developed low-power and long-range communication device, so-called DatalogV1, provides automatic data gathering and reliable transmission. DatalogV1 incorporates self-monitoring for adapting measurement schedules for consumption management and to capture events of interest.

Two tests are used to assess the success of the development. The results show an autonomous and robust monitoring system for long-term collection of water level data in many sparse locations during ? ood events. Keywords: real-time data acquisition; sensor network; hydrological monitoring; ? ood warning system Sensors 2012, 12 1. Introduction 4214 A warmer climate, with its increased climate variability, will increase the risk of both ? oods and droughts [1], whose management and mitigation are important to protect property, life, and naturalenvironment. Real-time accurate monitoring of hydrologic variables is key for ? od forecasting, as well as for optimizing related warning systems for damage mitigation. Recent studies show that in the speci? c case of semiarid and arid areas, adequate deployment of monitoring networks is essential to a real understanding of the underlying processes generating run-off in storm events, and to achieve effective emergency systems (e. g. , [2]). Traditionally, researchers have directly collected data at the places of interest. This has now been commonly substituted by automatic sensor and datalogger systems, which provide high temporal data resolution, while reducing operational human resource requirements.

Dataloggers permit local automatic and unattended data gathering, and reduce environmental perturbation. However, data retrieval from standard dataloggers and storage in processing and control/warning centers still has to be done either manually, which prevents its applicability in ? ood warning systems, or through wired connections, which leads to substantial investments and operational costs. To confront these problems, sensor networktechnologyhas been proposed in many monitoring applications [3]. Yet, speci? c literature on sensor network for ? ood forecasting is sparse, with only a few examples available (e. . , [4–8]). Basically, a sensor network comprises a set of nodes, where each node includes a processor, a wireless radio module, a power supply, and is equipped with sensor hardware to capture environmental data. Each node performs the tasks of data gathering, physical parameter processing, and wireless data transmission to the control server. Speci? cally, for hydrologic applications, sensor nodes must also ful? ll a number of additional requirements: • Power lifetime: Power sources are often not available at the locations of hydrological interest.

Moreover, these locations are usually unprotected, and if renewable energy devices are used, there are prone to vandalism or theft. Thus, sensor nodes must have low-consumption, which along with existing standard batteries, should last at least one hydrologic cycle. • Sensor hardware compatibility: Most hydrologic sensor nodes include a datalogger device connected through a cable to one or more measurement instruments. The datalogger must provide multiple wired interfaces to be able to communicate with a range of speci? c sensor hardware interfaces.

This also involves issues of power supply, and selective time for power dispatching, which leads to optimal power management and facilitates the expansion of connected instruments. • Reliability: Harsh weather conditions may cause failures in the wireless communication over the monitoring network. Backup mechanisms in local sensor dataloggers must be used to avoid information losses in unexpected crashes. • Long-range communication: Hydrologic measurement locations are commonly sparse over large areas, and far away from the control center (i. e. , tens or hundreds of kilometers).

Sensor nodes must have a peer-to-peer connection with the control center. Sensors 2012, 12 4215 In general, these, sometimes opposing, requirements are dif? cult to be satis? ed by existing developed solutions. For example, multiple sensor readings and long-range communication are high power-consumption tasks, which diminish battery lifetime. For instance, many existing wireless solutions for agriculture applications (e. g. , [9–11]) use a set of tens or hundreds of motes, which collaborate to gather dense data in a small area. Motes have low consumption, but they provide limited sensor interfaces, and short-range communication.

On the other hand, several hydrologic and meteorologic applications have been implemented with a few wireless datalogger stations, which individually obtain multi-sensor data in a few sparse locations over a large area (e. g. , [5, 12–14]). These dataloggers permit high computing and long-range communication. However, they have an excessive investment cost and a high consumption that may be, in the long-term, unsustainable. This paper describes the design, development, and deployment of a real-time monitoring system for hydrological applications.

The paper is focused on the description in detail of our wireless datalogger device, so-called DatalogV1 [15], which combines the low consumption of motes and the reliable communication of most powerful multi-sensor datalogger stations in order to satisfy the requirements of ? ood warning system scenarios. The DatalogV1 provides automatic monitoring and long-term autonomy in sparse points over large areas. To demonstrate the goodness of the DatalogV1 design, we deployed a monitoring network in the Rambla del Albuj? n watershed, in Southern Spain. The severity of ? ash ? ods in the Rambla del o Albuj? n has caused important environmental and economic damages over the last years. Accordingly, the o wireless monitoring network is intended to provide real-time accurate hydrologic information to support an operational model-based ? ood warning system. This is an excellent test to asses the DatalogV1 performance and success in a real case scenario. The remainder of the paper is organized as follows. Section 2 introduces the context of environmental monitoring and ? ood warning systems. Section 3 depicts our hydrologic monitoring scenario.

Section 4 presents the design of DatalogV1 hardware. Section 5 shows the implementation of DatalogV1 software. Section 6 describes the architecture developed for remote hydrologic monitoring. Section 7 describes the deployment of the monitoring network in the Rambla del Albuj? n watershed. Section 8 shows the results o obtained regarding power consumption and data collection. Section 9 provides concluding remarks. 2. Environmental Monitoring Environmental monitoring is the most popular application for sensor networks. At present, sensor networks have been applied for a number of applications as, e. . , soil moisture monitoring [16], solar radiation mapping [17], aquatic monitoring [18], glacial control andclimate change[19], forest ? re alarm [20], landscape ? ooding alarm [21], and forecasting in rivers [22]. The ability to place autonomous and low cost nodes in large harsh environments without communication infrastructure enables accurate data collection directly observed from interest areas. With sensor networks, environmental data can be observed and collected in real-time, and used for forecasting upcoming phenomena and sending prompt warnings if required.

Sensors 2012, 12 2. 1. Model-Based Flood Warning System Context 4216 The developed sensor network was incorporated within the context of a model-based ? ood warning system in the Rambla del Albuj? n watershed. A model-based ? ood warning system, for mitigating the o effects of ? ooding on life and property, incorporates a catchment model based on observed/forecasted rainfall and telemetered observations of hydrologic state variables at various locations within the catchment area. Generally, observed variables are ? ow and/or water level in channels.

Also, other variables such as soil moisture and piezometric levels may be of interest, depending on the watershed response. Real-time updating of the ? ood forecasting involves the continual adaptation of the model state variables, outputs and parameters, so that the forecasts for various times into the future are based on the latest available information and are optimized, in some sense, to minimize the forecasting errors (e. g. , [23]). This is the process of data assimilation. Implementation of environmental sensor networks for data assimilation within model-based ? ood warning systems involves complex engineering and system challenges.

These systems must withstand the event of interest in real-time, remain functional over long time periods when no events occur, cover large geographical regions of interest to the event, and support the variety of sensor types needed to detect the phenomenon [8]. 3. Hydrological Monitoring and Forecasting in the Rambla del Albuj? n Watershed o The Rambla del Albuj? n watershed (650 km2 ) is the main drainage catchment in the Campo de o Cartagena basin, in Southern Spain (see Figure 1). The main channel in the watershed is 40 km long and ? ows into the Mar Menor; one of the big coastal lagoons in the Mediterranean (135 km2 ).

The Campo de Cartagena basin is an area with semiarid Mediterranean climate, where the average temperature ranges from 14 o C to 17 o C, mean potential evapotranspiration is 890 mm yr–1 and mean precipitation is 350 mm yr–1 . Most rainfall comes in short-time storm events, and the watershed hydrologic response is highly complex and non-uniform. Previous studies have shown the complex ? ash-? ood response of the Rambla del Albuj? n watershed o and the importance of spatially distributed observation for adequate forecasting (e. g. , [2]). Also, for ? ooding evaluations, stage gauges provide an advantage over ? w gauges that the observations remain unbiased when ? ow goes out of banks, in which case the validness of calibrated rating curves (stage-? ow relationships) is prevented. In this sense, remotely-sensed information (from aerial photography and/or satellites) is appealing as it contains much more spatial information than typical stage gauge networks in operational watersheds. Accordingly, recent studies are evaluating the potential of aerial photography and remotely sensed (from satellites) synthetic aperture radar to provide measurements over large areas of water levels and ? od extents in lakes and rivers (e. g. , TerraSAR-X or COSMO-Skymed constellations [24]). However, the current low temporal frequency of satellite acquisitions relative to gauging station sampling indicates that remote sensing still does not represent a viable replacement strategy for data assimilation into model-based forecasts [25]. Also, before the ? ow goes out of banks, the accuracy of standard stage gauges is higher than that provided by airborne information, which is key for early warnings.

Thus, if economically viable, a spatially distributed network of stage gauges remains the best option to capture the observations required to feed the forecasting and data assimilation processes. Sensors 2012, 12 4217 At the Rambla del Albuj? n watershed, we implemented a hydrological monitoring system consisting o on a network of stage gauges located at eight critical junction points between major tributaries. The monitoring locations were carefully chosen in order to achieve effective water level monitoring during ? ood events and a reliable model-based forecasting system.

Figure 1 shows the selected locations which are far away (? 50 km) from the control center at the University of Murcia, to the North of the watershed. In this area, an existing phone infrastructure enables the communication among the server in the control center and the DatalogV1s in the ? eld. The DatalogV1s must be autonomous only with batteries, because no power source exists in the monitoring area and solar panels are frequently stolen or vandalized. In the following sections, we describe the design and development of the DatalogV1 to provide remote data gathering of the water stage in channels during ? ods. Figure 1. Deployment scenario. The embedded image shows the location of the Rambla del Albuj? n watershed at the Southeast of the Iberian Peninsula. The violet line describes the o watershed boundary drawn on a digital terrain model (DTM). Within the watershed, the main channel network is shown in blue, and labeled squares indicate deployed gauge locations. Sensors 2012, 12 4. Design of DatalogV1 Hardware 4218 The DatalogV1’s design was developed to address the requirements of the described application. The block diagram of DatalogV1 is illustrated in Figure 2(a).

The critical components are a low-power microcontroller (µC) module that supervises the DatalogV1’s operation, multiple sensor interfaces (Pulse, SDI-12, RS-485, Analog) that enable to take measurements from different kinds of sensor devices, and a GPRS module for long-distance communication with the control center. Moreover, two communication modules (USB and Bluetooth) enable the in-situ interactions via a laptop. All electronic components and a battery are mounted in an IP65 waterproof box to protect from harsh weather conditions, as shown by Figure 2(b).

The DatalogV1’s design is balanced between low-power consumption for long-lifetime, and computational capability for multi-sensor reading and long-range communication. The hardware design of these components is described in the next subsections. Figure 2. Two different views of the DatalogV1. (a) Block diagram showing the main components. (b) The electronic components and the battery are mounted on a IP65 protection box. SDI-12 Interface RS-485 Interface Pulse Counters Analog Inputs Power Connector DC/DC Converter GPRS Module Linear Regulator Battery Connector Linear Regulator

Mosfet Switch µC DC/DC Converter Pulse Counters Bluetooth Module RS-485 Interface USB Module Battery Connector Power Connector Analogic Inputs SDI-12 Connector GPRS Module Bluetooth Module USB Module µC (a) (b) 4. 1. Design of Microcontroller Module The circuit schematic of the microcontroller module is shown in Figure 3. The central part of the schematic represents the low-power 8-bits microcontroller (PIC18LF8722) manufactured by Microchip. The PIC18F8722 operating to 3. 3 V is ideal for low power applications ( nanoWatts) with 120 nW sleep mode and 25 µW active mode.

It provides high processing speed (40 MHz) with a large 256 KB RAM memory. A 12 MB data? ash memory is included for local storage of sensor data. The top-left portion of the schematic (IC3) shows a security mechanism to avoid microcontroller blockage in case that available energy is not enough. Thus the microcontroller resets when there is less than 2. 4 V. The center-left part of the schematic contains the crystal oscillator setting to 11 MHz. (OSC1/OSC2 tags). The oscillator provides a precise clock signal to stabilize frequencies for sensor readings and data transmissions. Sensors 2012, 12 Figure 3.

Circuit schematic of the microcontroller module. The center portion is the microcontroller used to control DatalogV1 operation, and the center-left is the crystal oscillator used for setting the clock. 4219 4. 2. Design of Sensor Interfaces DatalogV1 provides multi-sensor interfaces to take readings from a wide set of hydrologic instruments. Its sensor interfaces are two pulse counters, two digital connectors (RS-485 and SDI-12), and eight analog inputs. Each pulse counter reads from a tipping-bucket rain gauge (pluviometer) which generates a discrete electrical signal for every amount of accumulated rainfall.

Digital interfaces supply power to and read measurements from instruments, which can themselves include some degree of computational capability. Analog connectors enable the reading of simple instruments which modify the supplying voltages to return voltage values proportional to the physical observed variables. These multiple interfaces are compatible with the most of hydrological sensor devices in the market. Pulse-counters typically connect to rain-gauge devices. The standard rain gauge collects the precipitation into a small container. Every time the container is ? led and emptied, it generates a electric pulse. According to the number of pulses and the size of the container, DatalogV1 estimates the precipitation without requiring power supply. Sensors 2012, 12 4220 For each digital interface, DatalogV1 can supply and read multiple sensors. Both RS-485 and SDI-12 interfaces consist of three electronic wires for data, ground and supplying voltage. The RS-485 is a standard serial communication for long distance and noisy environments. In addition, the SDI-12 is a serial data interface at 1, 200 baud designed for low-power sensors.

Using serial protocols, DatalogV1 can directly obtain the physical measurements. The analog inputs allow to read 8 differential sensors, 16 single-ended sensors, or a combination of both options. A differential connection comprises four electronic wires acting as voltage-supplier, ground, positive-voltage, and negative-voltage, while a single-end connection contains two electronic wires for supplying-voltage and positive-voltage. The main difference between differential and single-ended is the way to obtain the voltage value. In single-ended, the voltage value is the difference between the positive voltage and the ground at 0 V.

However, single-ended connections are sensitive to electrical noise errors, which are solved by differential connections. Because twisting wires together will ensure that any noise picked up will be the same for each wire, the voltage value in differential inputs is the difference between the positive and negative voltages. Figure 4. Circuit schematic of analog interfaces. (a) Selector of analog connections to plugged-in sensors, (b) ADC converter from output voltage to digital data. (a) (b) To obtain the measurements of the physical variables, output voltages are processed using three main hardware components: multiplexer, ampli? r, and ADC converter. Two multiplexers MC74HC4051D from Motorola company enable to select the output voltage of a speci? c analog sensor (Figure 4(a)). Each multiplexer contains 3 control pins CA0, CA1, and CA2 to choose an output voltage among 16 possibilities. The selected output voltage is ampli? ed for preserving high effective resolution. DatalogV1 uses an AD8622 ampli? er, manufactured by Analog Devices, that provides high current precision, low noise, and low power operation. The pre-con? gured ampli? cation depends on the output range Sensors 2012, 12 4221 of the selected sensor.

Finally, the ampli? ed output signal is converted to a digital value through an Analog-Digital Converter (ADC), as shown by Figure 4(b). DatalogV1 contains a 13-bit ADC MCP3302, manufactured by Microchip, that provides high precision and resolution. This ? exible design provides full compatibility with presumably all kind of available sensors for hydrologic use. 4. 3. Design of GPRS Communication Module A GPRS module is used to transmit monitoring data from DatalogV1 to the control center. Figure 5 shows the GPRS module implementing all functions for wireless communications. Figure 5.

Circuit schematic of the GPRS module. The center portion is the GPRS module used to control the long-distance communication, and the top-left portion is the SIM card connection. The top-left part of the circuit shows the connection of SIM phone-cards according to the manufacturer speci? cation. The bottom-left shows a uFL coaxial connector to the wireless antenna. We chose a Wavecom Q2686 chip, which is connected to the microcontroller via an USART interface (CS-USART). The Wavecom Q2686 contains a programmable 256 KB SRAM memory and includes a ARM9 32-bit processor at 104 MHz.

This Q2686 chip makes possible to join a GSM/GPRS base-station and receive/send data reliably in quad-band communications on the 800, 900, 1, 800 and 1, 900 MHz Sensors 2012, 12 4222 bands. Also, the chip makes it easy to upgrade to 3G when needed. This GPRS module enables long-distance UDP/IP communications through cellular radio networks. 4. 4. Design of Power Module The power module consists of two power sources and three regulable mechanism to provide a secure supply of electronics components. The main energy source is a 12 V DC battery of 7, 000 mAh power capacity which can be rechargeable using an optional solar panel.

To adapt the input tension of the solar panel (17–20 V) to a lower tension (12–15 V) to supply the battery, we use a commutated DC/DC regulator in step-down mode, as shown by Figure 6(a). The microcontroller turns on the DC/DC regulator when it detects that the battery has a low level according to a pre-established threshold. Three circuits guarantee stable energy levels for battery, solar-panel, and sensors, as shown by Figure 6(b). The circuits of battery and solar-panel include security mechanisms to avoid a too low power level input to the sensors.

For this, the circuit of sensors is used, before readings are taken, to check if the power supply is stable as to obtain an accurate measurement. Figure 6. Circuit schematic of the battery, solar-panel, and power-control modules. (a) Battery and solar modules, (b) secure power control for battery, solar panel, and sensor. (a) (b) Figure 7. Circuit schematic of the power supply module. (a) Power supply for GPRS, sensors, and ADC converter, (b) power supply for microcontroller. (a) (b) To reduce the power consumption, DatalogV1 keeps almost all electrical components deactivated, such as GPRS, sensors, and ADC.

Only the microcontroller circuit is always supplied at 3. 3 V Sensors 2012, 12 4223 (Figure 7(a)) through a linear regulator LM2936 from National Semiconductor with ultra-low current in the stand-by mode. This LM2936 regulator features low drop-out voltage (50 mA) to minimize power losses. Also, this circuit includes a diode (D10) to provide a security power to protect the microcontroller and all board at most 5 V. When it is necessary, the microcontroller supplies independently the electrical components using two DC/DC converters, two linear regulators and a MOSFET switch (Figure 7(b)).

Concretely to supply sensors, a DC/DC converter and the MOSFET switch is combined to create a adjustable commutation cell. The design of the commutation cell includes high-power isolated chips in order to reduce interferences. At the same time, it has a good linearity and load regulation characteristics, and allows to establish the voltage supply between 3 V and 10 V. The chosen MOSFET is a FDC6330L, manufactured by Fairchild Semiconductor, which provides high performance for extremely low on-resistance (